)]}'
{
  "commit": "a0043c6d3027339eec6db415a88af975e0274bab",
  "tree": "81ed10ebc70dafcefbb5ebd77197fbbbc616522b",
  "parents": [
    "fc981147618756b5f02fc64d987681600bd3d5fe"
  ],
  "author": {
    "name": "Yunqing Wang",
    "email": "yunqingwang@google.com",
    "time": "Wed Mar 25 14:19:29 2015 -0700"
  },
  "committer": {
    "name": "Yunqing Wang",
    "email": "yunqingwang@google.com",
    "time": "Wed Apr 01 09:43:40 2015 -0700"
  },
  "message": "Enhance the transform skipping decision-making in non-rd mode\n\nFor large partition blocks(block_size \u003e 32x32), the variance\ncalculation is modified so that every 8x8 block\u0027s variance\nis stored during the calculation, which is used in the\nfollowing transform skipping test. Also, the variance for\nevery tx block is calculated. The skipping test checks all tx\nblocks in the partition, and sets the skip flag only if all tx\nblocks are skippable. If the skip flag of Y plane is 1, a\nquick evaluation is done on UV planes. If the current partition\nblock is skippable in YUV planes, the mode search checks fewer\ninter modes and doesn\u0027t check intra modes.\n\nThe rtc set borg test(at speed 6) showed that:\nOverall psnr: -0.527%; Avg psnr: -0.510%; ssim: -0.573%.\nAverage single-thread speedup on rtc set was 3.5%.\nFor 720p clips, more speedups were seen.\ngipsrecmotion: 13%\ngipsrestat: 12%\nvidyo: 5 - 9%\ndark: 15%\nniklas: 6%\n\nChange-Id: I8d8ebec0cb305f1de016516400bf007c3042666e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "f8a5e6ae7ba9618dda8ad6234cb70e85e2f13fd2",
      "old_mode": 33188,
      "old_path": "vp9/encoder/vp9_pickmode.c",
      "new_id": "860716209b2ee00e5289d4c70a2a01611857e2e7",
      "new_mode": 33188,
      "new_path": "vp9/encoder/vp9_pickmode.c"
    }
  ]
}